If you've ever studied single-variable calculus, you know how to find the derivative of f(x) — the rate of change of f as x varies. But what happens when your function depends on two or more variables, like f(x, y) = x² + y²? That's where partial derivatives come in.

In one sentence: A partial derivative measures how a function changes when you tweak one variable at a time, while keeping all other variables frozen.

The Core Intuition

Imagine you're standing on a mountain (a 3D surface defined by f(x,y) = altitude). If you walk east, your altitude changes at a certain rate — that's ∂f/∂x. If you walk north, you get a different rate — that's ∂f/∂y. The partial derivative isolates one direction of change at a time.

Mathematically, we hold all other variables constant and differentiate as if we have a single-variable function. The notation uses a curly "∂" instead of the straight "d" to signal that other variables exist but are being frozen.

The Formal Definition

The partial derivative of f(x, y) with respect to x is defined as the limit:

∂f/∂x = lim[h→0] [f(x + h, y) − f(x, y)] / h

This is identical to the single-variable derivative definition — the only difference is that y appears in both f(x+h, y) and f(x, y) and never changes. When we compute this limit, y is just a constant parameter.

How to Compute Partial Derivatives

The good news: you already know all the differentiation rules. Power rule, product rule, chain rule — they all apply. The only new skill is recognizing what's a constant.

Rule: Treat other variables as constants

Let's work through f(x, y) = x³y + sin(y)x + y².

To find ∂f/∂x, treat y as a constant:

∂f/∂x = 3x²y + sin(y) · 1 + 0 = 3x²y + sin(y)

Note: y² differentiates to 0 because y is treated as a constant with respect to x.

To find ∂f/∂y, treat x as a constant:

∂f/∂y = x³ · 1 + cos(y) · x + 2y = x³ + x·cos(y) + 2y

Try it yourself: Use our partial derivative calculator to verify these results. Enter x^3 * y + sin(y) * x + y^2, select variable x or y, and click Calculate.

Second-Order Partial Derivatives

Just as in single-variable calculus, you can differentiate again. Second-order partial derivatives appear in Taylor series, the Hessian matrix, and wave equations.

For f(x, y) = x²y³:

∂f/∂x = 2xy³ ∂²f/∂x² = 2y³ (differentiate again w.r.t. x) ∂f/∂y = 3x²y² ∂²f/∂y² = 6x²y (differentiate again w.r.t. y)

Mixed Partial Derivatives

You can also differentiate with respect to x first, then y — or vice versa. These are called mixed partials:

∂²f/∂y∂x = ∂/∂y (∂f/∂x) = ∂/∂y (2xy³) = 6xy² ∂²f/∂x∂y = ∂/∂x (∂f/∂y) = ∂/∂x (3x²y²) = 6xy²

Both give the same answer! This is Schwarz's (Clairaut's) Theorem: mixed partial derivatives are equal when both are continuous. The order of differentiation doesn't matter.

Common Notation

Partial derivatives appear in several notations. You'll encounter all of them:

  • Leibniz notation: ∂f/∂x, ∂²f/∂x², ∂²f/∂x∂y
  • Subscript notation: fₓ, fᵧ, fₓₓ, fₓᵧ
  • Operator notation: ∂ₓf, ∂ᵧf

The subscript notation (fₓ meaning ∂f/∂x) is common in textbooks and is compact. The Leibniz notation (∂f/∂x) is more explicit and less ambiguous — it's what our calculator uses.

The Gradient: All Partials Together

When you collect all partial derivatives into a single vector, you get the gradient:

∇f(x, y, z) = ( ∂f/∂x, ∂f/∂y, ∂f/∂z )

The gradient is a vector that always points in the direction of steepest increase of the function. Its magnitude tells you how steep the slope is in that direction. This makes it indispensable in optimization.

Real-World Applications

Machine Learning — Gradient Descent

Every time a neural network trains, it computes partial derivatives of the loss function with respect to every weight and bias. The algorithm then moves the weights in the direction opposite to the gradient — nudging them to minimize loss. Without partial derivatives, modern AI would be impossible.

Thermodynamics

State functions in thermodynamics — like internal energy U(T, V, n) — depend on temperature, volume, and amount of substance simultaneously. The Maxwell relations are statements of equality between mixed partial derivatives of thermodynamic potentials.

Economics — Marginal Analysis

In economics, if profit Π depends on price p, quantity q, and advertising spend a, then ∂Π/∂p tells you the marginal effect of a price change. This is crucial for setting prices and allocating budgets.

Fluid Mechanics

The Navier–Stokes equations — governing fluid flow — are a system of partial differential equations (PDEs). Every term is a partial derivative of velocity or pressure with respect to space and time.

Common Mistakes to Avoid

  • Forgetting the chain rule: If x appears inside a composite function like sin(x²y), you need the chain rule even when differentiating w.r.t. x.
  • Treating all letters as variables: If a problem says a and b are constants, differentiate them to zero — even if they look like variables.
  • Confusing ∂ with d: Never write d/dx when you mean ∂/∂x for a multivariable function. They are different operations conceptually (though the computation is similar).
  • Forgetting products: In ∂(x·y)/∂x, y is constant so the answer is just y — not 1. You're not differentiating y.

Putting It All Together

Partial derivatives extend the power of differential calculus to functions of multiple variables. The core skill is simple: differentiate with respect to one variable while treating others as constants. From there, you can build toward the gradient, the Hessian matrix, Lagrange multipliers, and partial differential equations.

The best way to build fluency is practice. Our free partial derivative calculator lets you verify your work and explore examples instantly — complete with step-by-step solutions so you can follow the reasoning.

Ready to practice? Head to the calculator and try computing ∂/∂x of x^2 * sin(y) + e^(x*y). Check your answer step by step!

WRITTEN BY
PartialDeriv Team
Calculus Multivariable Beginner Tutorial
∂

Ready to Compute?

Try our free partial derivative calculator — step-by-step solutions included.

Open Calculator →